mirror of
https://github.com/nlohmann/json.git
synced 2026-09-24 06:10:22 +01:00
claude/iterative-diff
3
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 SamareshSingh
|
7192763f15 |
Fix compile error when using nlohmann ordered_map with WITH_DEFAULT macros (#5163)
* Fix compile error when using nlohmann ordered_map with WITH_DEFAULT macros ordered_map inherits its copy and move assignment from the underlying std vector, which requires value_type to be CopyAssignable. value_type is pair<const Key, T> whose assignment is deleted because of the const Key, so any code that assigns ordered_map (for example the ternary in NLOHMANN_JSON_FROM_WITH_DEFAULT) fails to compile (issue #5122). Provide assignment operators on ordered_map that rebuild via clear plus push_back for copy and transfer the underlying buffer for move, neither of which needs pair assignment. Also switch the map-shaped from_json overload from a transform plus inserter idiom to a range-for plus emplace, which avoids the same hazard. Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> * Update ordered_map.hpp removed unwanted comments Signed-off-by: SamareshSingh <97642706+ssam18@users.noreply.github.com> Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> * Update json.hpp Signed-off-by: SamareshSingh <97642706+ssam18@users.noreply.github.com> Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> * Address CI issues for ordered_map fix Declare an explicit defaulted destructor on ordered_map so the rule of five is complete (clang-tidy cppcoreguidelines-special-member-functions and hicpp-special-member-functions). Initialize the ordered_map field in the regression test struct so GCC effective-C++ stops flagging Example_5122 with a missing member initializer. Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> * Suppress redundant-member-init lint on Example_5122::c The empty brace-init on c{} is required by GCC -Weffc++ to mark the member as initialized in the synthesized default constructor, but clang-tidy readability-redundant-member-init flags the same line because ordered_map already has a default constructor. The two checks pull in opposite directions, so add a targeted NOLINT to keep both happy. Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> * Address review: strong exception safety in copy-assign, simplify move-assign noexcept Copy assignment now constructs a temporary copy before move-assigning the Container subobject, preserving *this if the copy throws. Move assignment uses std::is_nothrow_move_assignable<Container> for a cleaner noexcept specifier, matching the style of the move constructor. Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> * Restore self-assignment check in copy-assign to satisfy cert-oop54-cpp clang-tidy's cert-oop54-cpp flagged the previous revision because it could not recognize the implicit self-safety of the copy-then-move pattern. Restore the explicit `if (this != &other)` guard — strong exception safety is preserved since the temporary copy is still constructed before the move-assign of the Container subobject. Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> * Address review: add explicit self-assignment and move-assignment tests for ordered_map Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> * Address review: gate -Wself-assign-overloaded suppression on Clang version -Wself-assign-overloaded was introduced in Clang 7. Older Clang versions fail the build with "unknown warning group" when the suppression pragma references it unconditionally. Use __has_warning inside an __clang__ branch so the suppression is only emitted on Clang versions that recognize the warning. The inner check stays inside the __clang__ guard because GCC does not provide __has_warning and would tokenize-error on the argument list. Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> * Address CI: drop unused gating macro to silence -Wunused-macros The previous attempt defined JSON_TEST_5122_SUPPRESS_SELF_ASSIGN_OVERLOADED as 0 unconditionally and then overrode it to 1 on Clang versions that recognize the warning. On those Clangs the initial define is immediately undef'd without being read, which trips Clang's -Wunused-macros under -Weverything in the ci_test_clang job. Drop the macro and gate the DOCTEST_CLANG_SUPPRESS_WARNING_PUSH/POP pragmas directly with __has_warning inside the existing __clang__ branch. Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> --------- Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> Signed-off-by: SamareshSingh <97642706+ssam18@users.noreply.github.com> |
||
|
SamareshSingh
|
62f3b41b30 |
fix: treat single-element brace-init as copy/move instead of wrapping in array (#5074) (#5090)
* fix: treat single-element brace-init as copy/move
When passing a json value using brace initialization with a single element
(e.g., `json j{someObj}` or `foo({someJson})`), C++ always prefers the
initializer_list constructor over the copy/move constructor. This caused
the value to be unexpectedly wrapped in a single-element array.
This bug was previously compiler-dependent (GCC wrapped, Clang did not),
but Clang 20 started matching GCC behavior, making it a universal issue.
Fix: In the initializer_list constructor, when type deduction is enabled
and the list has exactly one element, copy/move it directly instead of
creating a single-element array.
Before:
json obj = {{"key", 1}};
json j{obj}; // -> [{"key":1}] (wrong: array)
foo({obj}); // -> [{"key":1}] (wrong: array)
After:
json j{obj}; // -> {"key":1} (correct: copy)
foo({obj}); // -> {"key":1} (correct: copy)
To explicitly create a single-element array, use json::array({value}).
Fixes the issue #5074
Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com>
* fix: regenerate amalgamated single_include/nlohmann/json.hpp
- Add missing comment from include/nlohmann/json.hpp explaining the
single-element brace-init fix (issue #5074)
- Fix extra 4-space indentation in embedded json_fwd.hpp section
Regenerated by running: make amalgamate
Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com>
* Revert brace-init semantics change and fix amalgamation
The single-element brace-init change was a breaking change that cannot be accepted upstream. Reverted all related source, test, and doc changes, then regenerated single_include with correct indentation to pass the amalgamation CI check.
Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com>
* Fix: add JSON_BRACE_INIT_COPY_SEMANTICS opt-in macro for issue #5074
Single-element brace initialization wrapping in an array cannot be fixed without breaking existing code. Added JSON_BRACE_INIT_COPY_SEMANTICS as an opt-in macro (default 0) so users can enable copy/move semantics for single-element brace init without affecting anyone relying on the current behavior.
Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com>
* docs: add dedicated macro page and CI test target for JSON_BRACE_INIT_COPY_SEMANTICS
Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com>
* fix: remove compiler-dependent assertions from #5074 regression test
Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com>
* fix: use defined() guard for JSON_BRACE_INIT_COPY_SEMANTICS to satisfy -Wundef
Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com>
* docs: fix section name in json_brace_init_copy_semantics.md to pass style check
Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com>
* docs: move Default definition section before Notes to fix style check order
Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com>
---------
Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com>
|
||
|
SamareshSingh
|
0422165315 |
Handle 'moved' events in serve_header.py (#4997)
Fixes #3659 I was testing serve_header.py with my local development setup and noticed that when I moved directories into or out of the monitored root, they weren't being picked up properly. The script would only detect create and delete events but not move operations. This was happening because the on_any_event handler only checked for 'created' and 'deleted' events on directories. Move events have a separate event type 'moved' that includes both the source and destination paths. The fix treats a move event like a combination of delete (for the source) and create (for the destination) - we rescan to remove any trees that were moved out, and add the destination directory to check for new trees that were moved in. This should make the development workflow smoother when reorganizing project directories while the server is running. Signed-off-by: Samaresh Kumar Singh <ssam3003@gmail.com> |